Multicentre validation and comparison of 23 noninvasive diagnostic models for liver fibrosis in treatment-naïve Chinese patients with autoimmune hepatitis

BACKGROUND: Limitations of liver biopsy have driven the development of noninvasive liver fibrosis assessment models. Most noninvasive liver fibrosis models have been developed in hepatitis B or C cohorts but not in autoimmune hepatitis (AIH) cohorts. This study aimed to evaluate and compare the predictive performance of 23 existing noninvasive models for fibrosis staging in untreated Chinese patients with AIH. METHODS: A multicentre retrospective study was conducted on 258 untreated patients with autoimmune hepatitis who underwent liver biopsy. Histological fibrosis stage was evaluated using the Ishak scoring system. The diagnostic accuracy of the 23 noninvasive models was assessed by the area under the receiver operating characteristic curve (AUROC), including the aspartate aminotransferase (AST) to alanine aminotransferase (ALT) ratio (AAR), age-platelet index (APind), aspartate aminotransferase (AST) to platelet ratio index (APRI), cirrhosis discriminant score (CDS), Doha score, fibrosis-4 index (FIB-4), fibrosis quotient (FibroQ), globulin/platelet model (GP), gamma-glutamyl transpeptidase to platelet ratio (GPR), Göteburg University cirrhosis index (GUCI), Hb-F, King's score, Lok index, Pohl score, S index, fibrosis-5 index (FIB-5), red cell distribution width-to-platelet ratio (RPR), Forns index, Fibro-alpha, globulin-albumin ratio (GAR), and age platelet count index. RESULTS: The median age of the 258 patients was 53 (45-58) years, in which 236 (91.47%) patients were females. Significant fibrosis and cirrhosis were present in 79.84% and 9.30% of patients, respectively. For predicting significant fibrosis, the AUROCs of APind, GP, FIB-4, HB-F, the Lok index, the RPR, the Fibro-alpha index, and the age platelet count index exceeded 0.700. All the models had AUROCs < 0.700 for advanced fibrosis, whereas the Doha score, FI, GP, RPR, and Fibro-alpha achieved AUROCs > 0.700 for diagnosing cirrhosis. In terms of the grading system, the GP model outperformed other noninvasive models for assessing liver fibrosis in this cohort. CONCLUSIONS: In Chinese treatment-naïve AIH patients, the use of the GP model has moderate diagnostic value for liver fibrosis and cirrhosis. Most existing noninvasive models demonstrate suboptimal accuracy, especially for advanced fibrosis. Further development and validation of noninvasive AIH-specific models are warranted.